Samsung Electronics and Broadcom said on July 25, 2026 they have signed a memorandum of understanding to expand their strategic collaboration across memory, foundry and advanced packaging, in a partnership the two companies say will exceed an estimated $200 billion through 2030. The MOU was unveiled at the AI Summit at The Midway in San Francisco, with Jinman Han, head of Samsung Foundry, and Broadcom president and CEO Hock Tan on stage together.
HBM4 And HBM4E For Broadcom's Next AI Accelerators
On the memory side, Samsung will supply Broadcom with High Bandwidth Memory, including HBM4 and HBM4E, to feed the merchant silicon side of Broadcom's AI accelerator business. That is the same product line that already sells to hyperscalers via Broadcom's ASIC partnerships with the largest cloud providers, so long-term HBM commitments effectively lock in one of the biggest downstream buyers of Samsung's advanced memory just as HBM4E samples begin shipping.
2nm And Sub-2nm Foundry, Plus 2.3D/2.5D Packaging
On the foundry side, Broadcom will manufacture products including Wireless Broadband Communications silicon on Samsung's 2-nanometer and below process technologies at Pyeongtaek. The scope explicitly extends to 2.3D and 2.5D packaging built on the 2nm platform, giving Samsung's advanced packaging line a marquee AI customer at exactly the moment TSMC's CoWoS capacity remains the industry's most binding constraint. "AI is driving unprecedented demand for tightly integrated semiconductor technologies spanning memory, logic and advanced packaging," said Samsung DS Division vice chairman Young Hyun Jun.
A Foundry Escape Hatch From TSMC Dependency
For Broadcom, the deal is a supply-chain diversification play. TSMC still fabricates the vast majority of leading-edge AI logic, and hyperscalers building custom silicon with Broadcom have been asking for a credible second source for well over a year. Broadcom semiconductor solutions president Charlie Kawwas framed the MOU as "combining Samsung's memory and foundry expertise with Broadcom's AI and connectivity leadership." For Samsung, whose 2nm business has struggled to book anchor logic wins outside of Google and its own in-house Exynos program, an anchor customer with Broadcom's AI accelerator volume is exactly the demand signal Pyeongtaek needed to justify further capex.
Reshaping The 2nm Race
The MOU lands a week after Samsung Broadcom's $35B AI XPV platform with Apollo and Blackstone broadened the AI supply story, and after months of speculation that Broadcom was hedging away from a TSMC-only roadmap. Samsung's own AI compute stack, including its 1.3-trillion-won HBM expansion at Onyang, is being sized against exactly this kind of customer commitment. Analysts will now watch two things: whether the MOU converts into firm foundry tape-out schedules by early 2027, and whether Intel Foundry, still working through its own turnaround under Lip-Bu Tan, has anything left to counter with.
Reporting based on coverage from Samsung Global Newsroom, Bloomberg, CNBC, Manufacturing Dive and TrendForce.